INDEX

Explanations

locations mentioned with a high level of activation

oai_token-act-pair · gpt-3.5-turbo

instances of the word "at" used in various contexts

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

FTWARE

-0.76

alpha

-0.70

gravity

-0.69

Russ

-0.67

REDACTED

-0.65

HTTP

-0.63

istar

-0.62

Lens

-0.62

 WRITE

-0.61

PE

-0.61

POSITIVE LOGITS

 least

1.27

onement

0.98

abase

0.94

 halftime

0.84

rial

0.82

roph

0.80

 times

0.72

 liberty

0.71

oned

0.70

las

0.70

Activations Density 0.242%

locations mentioned with a high level of activation

instances of the word "at" used in various contexts

No Comments

No Known Activations